[m-rev.] for review: delete references to Erlang in the compiler

[Peter Wang]

On Wed, 28 Oct 2020 13:00:16 +1100 "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> 
> 2020-10-28 10:57 GMT+11:00 "Peter Wang" <novalazy at gmail.com>:
> > My own patch series was nearly ready as well, oh well. I didn't make the
> > change to item_type_repn_info so let's go with yours.
> 
> Sorry for stopping on your toes. I only started on the change after midnight,
> when it seemed to me that you stopped working on this diff series.
> 
> > The major difference with my patch series is that I kept lang_erlang
> > around. The compiler would continue to accept pragma foreign_code,
> > foreign_type, etc. for a language "Erlang" but silently discard them.
> > The compiler would also continue to accept --libgrades-exclude erlang.
> > My plan was that we would leave these transition measures in place for
> > one or two major releases (six or 12 months).
> 
> Sorry, but I don't see any point in that. Anyone who wants to keep using
> the Erlang grade would have to keep using pre-erlang-deletion
> compilers anyway, as fitting punishment for not speaking up on m-users.
> Anyone who has never used Erlang would not have any Erlang code
> to delete. The only people affected by the decision here are people
> who used Erlang in the past, but do not use it now.

My reasoning was that users may have imported a library that has Erlang
code, even if they aren't using it, e.g. the lex module or the
mercury-json module. Granted, those only have one foreign_proc each,
and they are probably representative, so it's not a big deal in practice.

> There is one thing we can do for users to help them, which is that
> when the compiler encounters "erlang" or "Erlang" as a foreign language
> on a foreign type, enum, code, decl or proc, the error message it emits
> should say not just "unknown language", but a message that specifically
> says that support for Erlang has been discontinued.

Ok.

Peter